Maybe the birdcage would be an easier place to start.
We began by placing the birdcage on the floor. I wanted Callie to sniff it out on her own terms. After a few minutes, she became bored and walked away. This was a good sign: she was getting used to it and didn’t view it as a threat. Next, I lay down on the floor and put my head in the birdcage. Callie still wasn’t ready to jump in with me. But a little peanut butter on my lips changed her mind. She bounded onto my chest and stuck her head in to lick it off.
Since she seemed to be having fun now, I dabbed a little peanut butter inside the coil to get her to go in by herself. She happily lapped it up. To avoid smearing the whole birdcage with peanut butter, I switched over to dog treats.
Each time she stuck her head in the birdcage, I moved the treat a little farther back. I wanted to see if she would assume the sphinx position in the birdcage, but I had no idea how to do that. As much as I loved Callie and secretly hoped that she was going to be subject number one, I was afraid that she was too ill behaved for the experiment.
I e-mailed Mark some pictures of the mock scanner. In the last photo, Callie lounged next to the head coil.
Much to my delight, it was Mark who suggested using her.
“She looks comfortable with it,” he wrote. “Why not make Callie the first subject?”
I test the mock head coil, while Callie investigates.
(Helen Berns)
9
Basic Training
CALLIE LOOKED GOOD AT HOME, but how would she do in an unfamiliar environment? She showed no fear of the head coil, a sign she would be able to adapt to novel tasks. But there was only one way to find out for sure.
Helen, eager to see how Callie would do with the training, helped me load her into the car, and the three of us headed to CPT with the head coil to see Mark work his magic.
Helen entered with Callie, while I placed the head coil on the floor.
Mark looked at it and nodded. “This should be easy. Did you bring treats?”
From puppy training, I knew that soft treats are best. You can cut them up into tiny pieces so the dog doesn’t fill up too quickly. And the dog can consume them easily without getting distracted by crunching on a hard biscuit. The only treats I could find around the house were some hot dogs that had been pushed to the back of the refrigerator. I had no idea how long they’d been there, but they smelled okay, and Callie loved them. I handed Mark a baggie full of sliced-up hot dogs.
“First,” he said, “let’s start with the clicker.”
A training clicker is a small device about the size of a USB flash drive that, unsurprisingly, makes a loud click when pressed. Dogs can hear the clicker from across the room. The advantage of using one is that it always makes the same sound, which is not the case with vocal commands. Because it’s almost impossible to screw up, the clicker is a useful tool for beginners like me. Its operation is simple: when the dog does something correct, you click. For this to work, however, you first have to teach the dog that a click equals a reward. This is classic classical conditioning. Just like Pavlov.
Callie tracked the bag of hot dogs as I handed it to Mark. Then she dutifully sat at his feet, tail sweeping the floor. Mark clicked and immediately gave her a piece of hot dog. Callie got even more excited. She could barely sit.
At this point, what Callie was doing was unimportant. Mark periodically clicked and handed her a reward. He was establishing the association of each click with a transfer of reward, making it a conditioned stimulus. It didn’t take long. A dozen click-rewards, and Callie understood the association. With the meaning of the clicker established, Callie was ready to learn a behavior. I could immediately see how the clicker was going to make this easier.
Mark explained another advantage of using the clicker. “We are going to shape her behavior. Initially, anything Callie does that is close to the desired behavior will be rewarded. The clicker makes it absolutely clear to her that she has done something correctly. This way, she won’t get conditioned to just my voice or your voice.”
The clicker gives instantaneous feedback, making it clear to a dog that she has done something good without wasting time fumbling for the treats. Unlike a human, a dog’s memory for what she has just done appears to be very limited. The longer the interval between the desired behavior and the subsequent reward, the less likely the dog will make the association. This phenomenon is called temporal discounting. Research in rats suggests that a reward given four seconds after a desired behavior is roughly half as effective as one given immediately. If the handler is deeply involved with the dog, using hand signals and vocal commands, he might not be able to give a reward for a while. This is especially true of complex behaviors. The clicker solves this problem by giving instantaneous feedback.
Mark was beginning to lure Callie into the head coil. Reaching into the coil with a hot dog in one hand and the clicker in the other, Mark had already succeeded in getting Callie to place her nose inside. Each time she did so, Mark clicked, praised her, and gave her a bit of hot dog.
With every click-reward, Mark pulled the food back a little bit, shaping her behavior gradually. Within ten repetitions, he had Callie crouching in the coil with her snout poking out the other end. Some gentle pressure on her rump indicated that she should lie down in the coil. As soon as she did, Mark clicked and exclaimed, “Good coil!” Callie wagged her tail and licked the hot dog from his hand.
I couldn’t believe how quickly Mark had gotten Callie where she needed to be.
“How is the positioning?” he asked.
Callie was lying down in a sphinx position in the coil. Her paws hung over the near edge. She would need to move back a little bit.
“We’ll want her head in the center.” Mark nudged her back an inch and clicked.
“You can shape her behavior at home too,” he said. “I think she’ll do really well with this.”
A woman walked into CPT with a border collie.
“This is Melissa Cate,” Mark said. “Melissa runs some of our agility classes at CPT. She’s interested in volunteering her dog for the MRI.”
“Mark told me about the Dog Project.” Pointing to her dog, she said, “This is McKenzie.”
McKenzie was Melissa’s three-year-old border collie. Melissa had begun agility competitions a few years earlier with her boxer, Zeke, who had reached the highest ranks. Zeke was now eight years old and slowing down a bit, so Melissa had gotten McKenzie as a puppy to keep competing in agility. They had been going strong ever since.
McKenzie was leggy and lean, about thirty-five pounds, with a long, thin head that would easily fit in the head coil. She trotted over to me and stared long and hard. She quickly realized that I was not a herdable animal and moved on to check out Helen.
Callie zoomed over and assumed a play bow with her front legs flat and her rump in the air, tail wagging like a vibrating string. We let the two dogs off-leash and they ran around the room. Callie did orbits around McKenzie, who seemed indifferent to the newbie dog.
It was time for McKenzie’s try with the head coil. With a dog treat, Melissa had no trouble coaxing her into the coil. Nibbling the food out of Melissa’s hand, McKenzie appeared unaware of the coil altogether. In agility competition, the dogs run through a serpentine tunnel, and McKenzie was completely comfortable in an enclosed space.
After a few minutes, Melissa commanded McKenzie to lie down. “Platz,” she said, using the German word for “down.” Mark explained that German words are commonly used in dog training because of the popular Schutzhund competitions. These began as training programs and tests for German shepherds but evolved into a full-fledged sport involving tracking, obedience, and protection phases.
With McKenzie lying down in the head coil, Melissa backed away to the other side of the room. McKenzie didn’t budge. In fact, she stayed motionless for a solid minute. When I saw what a well-trained dog like McKenzie could do, I knew we could really do this. If the dogs would go into the head coil, the
y would go into the MRI.
Melissa working with McKenzie in the head coil. Callie watches from across the room.
(Bryan Meltz)
So far, Mark and Melissa had been using basic behaviorist techniques. The appeal of behaviorism in dog training is its simplicity. By making rewards like food and praise contingent on desired behaviors, dogs quickly learn what to do to get something they want. But what do dogs think of this? After all, they aren’t robots bumbling around the world, randomly doing things and finding out which behaviors result in food. Dogs show purposeful and consistent behavior whether humans are there or not. This suggests that dogs have some internal mental model of how things work in their world. It is a limited model, of course. For instance, they don’t understand technology like computers or television. But dogs do understand how to get along with each other and with other species like humans, which is not an insignificant skill, and they don’t need treats to learn how.
While McKenzie remained still in the head coil, Callie watched with rapt attention. It’s possible she was interested in only the food being handed out, but her gaze wasn’t always tracking Melissa’s hands. Callie’s eyes darted back and forth between McKenzie and Melissa. You could almost see the wheels turning inside Callie’s head as she tried to figure out what was going on.
Even though we were using basic principles of behaviorism like positive reinforcement and shaping to train the dogs to enter and lie still in the head coil, Callie made it clear that a different type of learning was going on too. She was learning by observation.
Social learning, or imitation, is an obvious feature of human behavior. We can learn a great deal simply by observing what other people do. Strangely, dogs have not been given much credit for being able to do this too. But Callie illustrated clearly what everyone with more than one dog knows: dogs learn from each other.
Although behaviorist experiments dominate the canine research literature, there have been a few experiments demonstrating social learning between dogs. An old experiment found that puppies that observe littermates pulling a cart by a string can copy that behavior. Another study showed that puppies that watched their mother, a police dog, search for narcotics did better when learning this task compared to puppies that did not observe their mother first.
Relatively little is known about the neurobiology of social learning or imitative behavior. Even in humans, we don’t know much about which parts of the brain are involved. Whether dog or human, social learning is not contingent on rewards. So why does the puppy copy its littermates and pull the cart? After all, there is no food to be gained. Maybe the Dog Project would provide the answers.
But first, we needed to train Callie and McKenzie to perform a complicated task under noisy conditions.
Over the next few weeks, Callie and I worked with the head coil at home on a daily basis. I limited the sessions to ten minutes. Mark had explained that short, daily training is much more effective than infrequent long sessions. This prevents both the dog and the handler from getting bored. Consistency is the key.
Callie learned quickly. As soon as I reached for the head coil, Callie would jump up and down to try to get in it. Once inside, she assumed the sphinx position and waited for me to give her hot dogs. Her tail never stopped wagging.
“Good coil!” I would praise. More tail wags.
The next step was to introduce the chin rest. When we scan humans, the subject normally lies on his back with his head surrounded by foam padding in the head coil. The padding makes it comfortable while also preventing head movement. But the human setup wouldn’t work for a dog. Callie and McKenzie would have to lay on their stomachs, and I doubted that either of them would want her head surrounded by foam, like Leonard did with his monkeys at Yerkes.
I didn’t yet know how we were going to solve the head movement problem. The first step, though, was to give the dogs something they could rest their heads on. Something firm, yet comfortable. My first thought was the foam used in seat cushions, so I picked up the firmest foam I could find at the fabric store, cut off a chunk, and, while Callie was relaxing on the sofa, gently wedged it under her chin. She just let her head sink into it and went to sleep. That was a good sign, but the foam compressed too much to offer enough support. We needed something firmer.
I went to the hardware store and looked at foam insulation. Too hard. I was beginning to feel like Goldilocks. For several days, I searched in vain for something that would work. The chin rest would have to span the diameter of the foot-wide head coil. At that length, furniture foam collapsed with the slightest pressure in the middle of the span. But the materials at the hardware store were uncomfortably stiff.
The solution caught my eye in a sporting goods store. Helen and Maddy had wanted some new sneakers. While they were trying them on, I milled about, not looking for anything in particular. It was the middle of December, and the store was having a sale on summer swim gear. A stack of boogie boards was pushed into a corner with a handwritten sign: $5 each. I squeezed one. Firm, but not hard.
I took the whole stack to the register. The cashier looked at me like I was crazy.
“Science project,” I said.
At home, I used a utility knife to cut a strip of boogie board to match the inner diameter of the head coil. It made a comfortable bridge that didn’t collapse in the middle under pressure.
I got my baggie of hot dogs and approached Callie with the foam bar. She saw the treats and started wagging her tail. With her in the sphinx position on the floor, I gently pushed the foam under her chin.
“Touch,” I said, and gave her a treat.
As usual, Callie was a quick study. After a few repetitions, her head relaxed as soon as I gave the “touch” command, and I could feel the weight of her head against the foam bar. Pretty soon, I didn’t even have to cue her by pressing the boogie board against her chin. With the bar a centimeter below her chin, she would drop her head to make contact on command.
The next day, we practiced the “touch” command with Callie in the head coil. She got it. With the foam rest spanning the diameter, Callie scooted in and stuck her paws beneath it. I said, “Touch,” and she plunked her head on the bar.
“Good girl!” I exclaimed. She just wagged her tail. I couldn’t believe how quickly she was picking this stuff up. Lyra, drooling nearby, soon learned that she too would get hot dogs by hanging around the head coil. Lyra would then start barking if she didn’t get a piece of the action. She had a sensitive stomach, though, and tended to burp if I gave her too many hot dogs.
Each day we practiced with the chin rest in the head coil, and each day Callie held her head in position for longer and longer stretches. After a week of daily sessions, I no longer needed to say the command. I would just put the head coil on the floor, and she would go right in and plop her head on the rest.
We were gradually making the task more complex for Callie and McKenzie by adding elements before the final behavior. The technique is called backward chaining. Once Callie had learned to go into the coil, I placed the coil inside the mockup of the MRI bore. Since Callie knew that she would be rewarded only for going into the head coil, she trotted into the tube and into the coil, after which I promptly rewarded her. Next, I raised the tube to the height of the patient table of a real MRI machine. Callie would have to be taught to go up a set of doggie steps. These were designed for dogs to walk up to the height of their owners’ beds. Since they were made entirely out of plastic, they would be safe for use next to the real MRI.
It took a couple of days to teach Callie to go up the steps. I started by placing a hot dog on each step. Callie followed the trail of meat right to the top, where I gave her excessive praise. Once she was used to the steps, I placed them in front of the elevated tube and continued the meat trail all the way to the head coil. Once she was in the tube, I ran around to the other end and pointed to the head coil. She scooted in and waited for more treats.
The last and most challenging element was the scanner n
oise. Andrew had already recorded the jackhammer-like sounds of the MRI in action. Initially, we focused just on getting Callie and McKenzie used to the level of ambient noise. Later, we would need to figure out the exact scanner settings for the dogs, which would result in slightly different sounds.
I started by simply playing the scanner noise at low volume through a stereo while I did the training with Callie. She quickly learned to ignore it. Each day I would increase the volume a little bit. Soon, though, it would get to a level that was unpleasant. Time to introduce Callie to the earmuffs.
Callie sporting the earmuffs and learning to use the boogie board chin rest in the head coil.
(Gregory Berns)
Human subjects wear earplugs, but I had yet to meet a dog that would let you put anything in his ears. Plus, the ear canal of a dog makes a right-angle bend. If an earplug became wedged in the canal beyond the bend, we might not be able to get it out. The only alternative was earmuffs that went over the outside of each ear. Amazingly, I discovered Safe and Sound Pets, a company that makes Mutt Muffs. The founder of the company, a pilot of small aircraft, realized the need to protect his dog’s hearing when taking him up for flights. He adapted human earmuffs to a more triangular shape that would fit most dogs’ heads. We ordered several sets in different sizes.
Callie was not so thrilled with the Mutt Muffs. It took many bits of hot dog, first for putting her nose through the loop formed by the chin strap and eventually for letting me push the earmuffs back over her head. Even then, she would paw them off right away. I didn’t push her. With Mark’s advice, I gradually lengthened the time she had to wear them before giving her a piece of hot dog. Pretty soon, Callie would leave the earmuffs on, trot up the steps into the tube, and place her head on the chin rest in the head coil.
How Dogs Love Us: A Neuroscientist and His Adopted Dog Decode the Canine Brain Page 8